Autonomous vehicle validation - are we just guessing or can we predict the future?
conference contributionposted on 2018-08-08, 08:50 authored by Steven Reed, Laurie BrownLaurie Brown
At present there exists no standardised, systematic or structured methodology for the validation of autonomous pod vehicles. The lack of process is compounded by minimal real-world collision or near miss data covering this vehicle type. This gap in knowledge may present a significant obstacle in autonomous pod vehicle development and consumer uptake from two sides; firstly, the scarcity of data for road safety practitioners to develop into appropriate test scenarios and secondly the lack of perceived transparency by the OEMs could form the perception that there is negligible vehicle development and manufacturer accountability, resulting in lack of trust from the end users. To counter this knowledge gap this paper aims to provide more information on initial steps into the definition of a systematic reference dataset which reflects both autonomous pod use and the capabilities of the vehicles to be tested. Due to the lack of real world data in the short term, it will be necessary to develop vehicle validation scenarios for autonomous pods based entirely on non-autonomous collisions of comparable vehicle types. Although there are currently no specific well-tested frameworks to follow, the approach discussed applies proven methodological principals from the field of general product design which assumes that if the design envelope is set correctly then any product that meets this, no matter how outlandish, will be valid.